How to get RDP classifier into .qza format?

[Nicholas_Bokulich]

Hi @microbiomeAnalyst,
The RDP classifier cannot be used in QIIME2.

The RDP reference database could be used as the reference database for training a classifier for use with classify-sklearn, however.

@microbiomeAnalyst

I have downloaded an aligned RDP taxonomy reference (Trainset16_022016.rdp.tgz) and would like to prepare this in a .qza format for DADA2 processing to be used for qiime2 taxonomic analysis.

You cannot process that file with dada2 — there are no quality scores! (I assume). dada2 is just for denoising.

@microbiomeAnalyst

On the data resources page there are classifiers trained on Silva and GreenGenes, but RDP is not one of the classifiers available for download.

There are lots of different reference databases out there, we can't list them all . To figure out how to import the RDP database, just check out the formats of those SILVA or Greengenes file — it should be straightforward to reformat the RDP reference database for your needs. Then follow this tutorial to import and train a classifier with those files.

@microbiomeAnalyst

In the documentation it states that the classify-sklearn method is essentially a python version of the popular RDP classifier, which is another Naive Bayes Classifier

Yep, both are naive Bayes classifiers — so same algorithm but slightly different implementations. For practical purposes you can consider these to be approximately the same (performance is rather similar, too).

@microbiomeAnalyst

Do I need to use this method to have a usable qza RDP classifier?

You need to use this method for any sklearn-based classifier. If that means training a naive Bayes classifier following the tutorial above, then yes, that is the method to use. You cannot actually use RDP classifier, i.e., the software package.

@microbiomeAnalyst

instead of the greengenes classifier I would to substitute the RDP classifier and then train this feature classifier (i.e. RDP classifier) to our 250 bases V3-V4 region.

Nope, cannot be done. You can use the same reference database, but need to use the sklearn classifier or one of the other taxonomy classifiers in QIIME 2 (e.g., the blast- or vsearch-based classifiers).

I hope that helps!